Two approaches for representing speech information with multichannel cochlear prostheses are being compared in tests with implant patients. Included in these studies are the compressed analog (CA) approach of a standard clinical device and research processors utilizing continuous interleaved sampling (CIS). Initial studies have been completed with nine subjects, seven of whom were selected on the basis of excellent performance with the Ineraid clinical processor, and the remaining two for their relatively poor performance with the same device. The tests include open-set recognition of words and sentences. Every subject has obtained a higher score--or repeated a score of 100% correct--on every test when using a CIS processor. These results are discussed in terms of their implications for processor design.